John Amuedo from Signal Inference gave a very interesting presentation and did a nice demo with the title of “SciCL: Lisp Extensions for Scientific Computing”. He showed how they processed a very bad copy of a famous movie ‘The Robe‘ (1953). Seeing Common Lisp as part of SciCL which was used for that kind of signal processing task was hilarious.
The following presentation was titled ‘Implementing a Parallelism Library for a Functional Subset of LISP’. The authors were David L. Rager, Warren A. Hunt Jr. from The University of Texas at Austin provided important insights about using Common Lisp for multicore CPU architectures.
‘AutoPrice: Automatic Unit Price Forecasting Utilizing Multiple Regression in Lisp’ which was presented by Erin Fleming from 2Is Inc. showed how Common Lisp was put to good use for a real world commercial data mining project, and a very interesting one indeed, which gave the company a competitive edge in their respective field.
Then it was our turn. I and my colleague Remzi Emre Başar presented our paper and software demo titled ‘Common Lisp for a Common Cultural Recommendation System: Creating a Cultural Recommendation & Exploration Engine Using Semantic Web’. It was very exciting to be in front of such an audience at MIT, about more than 200 Lisp hackers watching us. We had questions and valuable feedback which was very motivating.
Another important data mining / machine learning application was given by Robert P. Goldman, John Maraist from SIFT and LLC: ‘SHOPPER: Interpreter for a High-level Web Services Language’. I really liked the architecture they built for various ML learners and the flexibility of the system they have built using Common Lisp.
Following that, David Moon, one of the invited speakers talked about ‘Genuine, full-power, hygienic macro system for a language with syntax’. He laid down the fundamentals of a new language, PLOT (Programming Language for Old Timers).
The last presentation I have attended was by Joseph Marshall from Google, Inc. and its title was ‘An Unexceptional Implementation of Continuations’
The most exciting part of the day was ‘Debate: Are Macros a Menace?’ panel in which the advantages and disadvantages of the powerful Lisp macro system was discusses fiercely and cleverly.
The surprise of the evening was the unexpected lightning talk by Gerald Jay Sussman about the curriculum change in MIT and switching to Python from Scheme. Sussman talked about the changes in the concept of engineering, how it was perceived in 1980s and what changed in the late 1990s and how this affected programming in general. He said in 1980s they believed the best way to teach engineering was taking simple and very well understood little blocks / elements and combine them to create much more complex sophisticated systems. However, Sussman added, nowadays people were not trying to understand lots of details of lots of low-level systems, documentation, libraries, etc. but rather do natural science on the blocks, simply trying something and seeing if the result is okay and then using that component. Add some robotics stuff to that and for some core courses they switched to Python for introduction. (However that doesn’t mean Scheme / Lisp is totally dropped, on the contrary I have seen that Sussman is offering a very exciting course, ‘6.945 Adventures in Advanced Symbolic Programming‘ (Officially: Large-scale Symbolic Systems) which includes heavy duty coding assignments in Lisp).